Supported models and limits

Look up the AI models Autopilot supports and their context-window and output-token limits, organized by model.

Autopilot supports the following models. To configure your own subscription for any of them, refer to Configuring an LLM for Autopilot.

Supported models

Autopilot supports the following models:

• anthropic.claude-4-5-sonnet
• gpt-4o-2024-11-20
• gpt-5-1 (not available in Public Sector)
• gpt-5-4 (not available in Public Sector)
• gemini-2.5-pro (not available in Public Sector)
• gemini-2.5-flash (not available in Public Sector)

This list reflects the models Autopilot supports. For the model actually in effect for a given tenant and feature, and its live status (Active, Blocked, Deprecating soon), routing region, and configuration history, see Model hub.

Model limits

The following context window and maximum output token limits are specific to each model. They are separate from product-level and feature-level limits, such as maximum test cases or maximum requirement-quality suggestions — for those, see Best practices and limitations.

Model	Context window	Max output tokens
GPT-4o (OpenAI)	128K tokens	16,384 (16K) tokens
GPT-5 family (OpenAI)	400K tokens	Up to 128K tokens
Claude Sonnet 4.5 (Anthropic)	200K tokens	64K tokens
Gemini 2.5 Pro (Google)	Up to ~1M tokens (1,048,576)	~64K tokens (65,535)
Gemini 2.5 Flash (Google)	Up to ~1M tokens (smaller effective reasoning depth than Pro)	~64K tokens (65,535)